INFLATIONARY EXPECTATIONS AND U NCERTAINTY D URING THE
G REAT D EPRESSION IN G ERMANY
*
Hans-Joachim Voth
ABSTRACT
The unusual severity of the Great Depression in Germany has often
been interpreted as a unique failure of economic policy – with
disastrous consequences for the Weimar republic. In this paper, we
argue that criticisms of the Brüning government are misplaced. On the
basis of asset prices, we infer that fear of inflation was widespread in
1931/ 32, limiting the room for manoeuvre. We conclude that the
Britain's departure from gold – instead of facilitating such a step in
Germany – actually made it much more difficult to adopt
expansionary measures, because it increased uncertainty and fear of
inflation.
JEL classification codes
Keywords
E31, E43, E44, N14, N24
Inflation, Great Depression, Rational Expectations,
Uncertainty, Germany, Monetary Policy.
This version: December 6, 1999.
Email: [email protected]
*
Centre for History and Economics, Cambridge CB2 1ST, UK and Economics Department, Universitat
Pompeu Fabra, 08005 Barcelona, Spain. I would like to thank John James, Gavin Wright, and Tim Leunig for
their helpful discussion and conference participants at the All-UC conference in Santa Clara, California, March
1998 for their comments. The usual disclaimer applies.
2
Following the global financial crisis in the autumn of 1998, there has been widespread
1
concern about the dangers of deflation. Also, Japan's extended slump has led to renewed
2
interest in liquidity traps. Since the outbreak of the Asian crisis in 1997, the dilemmas of
economic policy in a deflationary environment are back on the agenda. Much of the
contemporary debate has turned to the Great Depression as the canonical case of how falls in
prices can get out of control, resulting in deflationary spirals. During the Great Depression,
prices were falling in all industrialized countries. Yet few economies experienced a more
pronounced deflation than Germany, where the consumer price index fell by 23 percent from
3
peak to trough. Using this celebrated case, we examine the extent to which policy choices in a
deflationary environment can be constrained by prior historical experience and simultaneous
decisions by other countries. In Germany, reductions of prices and wages were not simply a
result of weak demand and tight credit conditions; they were also part of a deliberate
deflationary policy, initiated by the Brüning government to combat the slump. Falling
nominal wages were supposed to increase the country's competitiveness on the world market,
acting as a 'surrogate devaluation'.
The new orthodoxy about the Great Depression argues that casting off the shackles of
the gold standard was crucial for recovery. Eichengreen's classic account details how
misconceived policy deepened the crisis and delayed the upturn in countries that remained on
4
gold, and how those countries that broke the link to gold quickly resumed growth. One of the
most successful examples of such a break with tradition was Britain's departure from gold in
September 1931. At the other end of the spectrum, Germany and France retained their
commitment to specie backing despite the deflationary consequences, balance sheet problems,
5
and financial uncertainty that it produced. In the case of Germany, one specific reason has
played a prominent role in discussions about policy options – fear of inflation. Prominent
scholars have argued that expansionary measures, and especially a departure from gold,
would have been impossible because memories of the hyperinflation continued to linger.
6
In order to establish if fear of inflation severely undermined the range of options for
government intervention, it has to be demonstrated that economic agents at the time did
1
Economist 1999.
Krugman 1999a, b.
3
For an overview, cf. Bernanke and James 1991.
4
Eichengreen 1992a. Cf. also Eichengreen and Sachs 1985.
5
Technically, since Germany suspended gold convertibility in 1931, it was off the gold standard – but did not
engage in any of the expansionary measures used in other countries until 1932/ 3. Cf. James 1985, p.
2
3
indeed fear price increases – despite the rapid fall in prices. We begin by placing the decline in
prices in Germany during the Great Depression in its historical context. This paper then seeks
to measure expectations of future price increases in Germany. To that end, we use a range of
indicators that allow us to reconstruct the price expectations implied by the behaviour of
economic agents. Next, we explore the consequences of the fear of inflation that began to
surface in the summer of 1931. The conclusion discusses implications for the political and
economic history of the Weimar Republic, and examines the issue of international policy coordination more closely.
FINANCIAL C RISIS AND D EFLATION
Did incompetent policymaking turn the German recession that started in 1928 into economic
7
– and ultimately, political – disaster? Numerous authors have claimed as much. Despite being
faced with an ever-deepening crisis, neither credit expansion nor deficit spending were used
on any significant scale before 1932. Some authors even suggest that the German government
8
deliberately exacerbated the crisis to liberate the Reich from the yoke of reparations. It is
therefore not surprising that, in recent accounts of the Great Depression, the German cabinet
under chancellor Brüning is often cited as an example of blatant mismanagement:
9
It is vivid testimony to the power of ideology that leaders like Brüning were induced to cling to
orthodoxy even as the world economy collapsed. He continued to advocate gold standard
policies after abandoning the gold standard himself. He ruined the German economy – and
destroyed German democracy – in the effort to show once and for all that Germany could not
pay reparations.
Apologetic interpretations have emphasized that any German government would have found
it hard to act decisively during the early 1930s, and that options for a more active policy were
10
largely conspicuous by their absence. Much of the literature on 'room for manoeuvre' during
the Great Depression in Germany, citing the recent memory of hyperinflation and
contemporary press comment, has argued that policymakers were constrained by fear of
inflation.
Germany could therefore not have followed Great Britain in leaving the gold standard,
even if it had wanted to, since the German public would have immediately anticipated a
6
Borchardt 1985.
Holtfrerich 1990, Kroll 1959, Bombach 1976.
8
Temin 1990, p. 82-3.
9
Feinstein, Temin and Toniolo 1997, p. 110.
10
Borchardt 1991, p. 146.
7
4
repeat of the inflationary experience 1919-23. Hence, the deflationary policies of the Brüning
government between 1930 and 1932 were not only rational, but without alternative. Being
unable to devalue directly, the only alternative was to let prices fall sufficiently to be
11
competitive on world markets. Expectations of price declines could only be reversed by the
price level falling sufficiently low to make it sensible to expect a rebound in the future. On July
th
24 1931, for example, Sir Horace Rumbold, the British ambassador to Berlin, wrote in his
th
report to the Foreign O ffice: "By 17 July the fear of inflation was very strong".
12
Historians
have spent considerable time and effort unearthing comments that attest to fear of inflation
13
during the early 1930s.
By December 1931, the German economist Wilhelm Lautenbach
was railing in a public lecture against what he saw as rampant 'inflation hysteria'.
14
Yet fearing price increases in a context of rapidly falling prices implies that
15
contemporaries were unable to grasp the nature of deflation.
At the peak of the deflation,
during the spring and summer of 1932, prices of industrial goods were falling at an
(annualized) rate of more than 15 percent. Once the price level had begun to weaken in 1928,
prices continued to fall until 1932. Even by the last quarter of 1935, prices had only regained
16
their Spring 1931 level. There is some indirect evidence that the nature of deflations was not
very clear to the wider public. Borchardt has shown how the term 'deflation' only came to be
used with any frequency in the popular press after 1931.
17
The 1928 edition of the popular
Brockhaus encyclopaedia defines deflation as a set of policy measures to reverse the price
18
increases during an earlier inflationary episode.
There is no appreciation of the fact that
deflationary pressures might arise as a result of economic shocks, nor any discussion of its
adverse affects.
11
Brüning referred to his deflationary policies as a 'surrogate devaluation'. Brüning 1970.
Cit. acc. to Borchardt 1985, p. 241. Cf. also Schulz 1985.
13
The most prominent statement in favour is Borchardt (1991, p. 150; 1985).
14
Borchardt 1985, p. 237.
15
Holtfrerich (1990, p. 69) also suggests that 'inflation' was simply a term used for any kind of expansionary
monetary or fiscal policy.
16
Using the price series for industrial finished goods ('industrielle Fertigwaren'), Konjunkturstatistisches Handbuch 1936, p. 104.
17
Borchardt 1985, figure 1, p. 238.
18
Brockhaus 1929, article 'Deflation', vol. 4, p. 463.
12
5
Despite evidence from diplomatic records and contemporary press reports, no
consensus has emerged – critics have pointed out that fear of inflation would have implied
patently irrational disregard for economic facts given the very considerable downward
momentum of prices in all sectors of the German economy. Fear of inflation might have been
clearly witnessed by some, but it is difficult to ascertain how wide-spread or deep it was. Some
observers have argued that fear of inflation was not always genuine and hardly common, and
was simply instrumentalized by certain politicians. For example, while the danger of inflation
was talked up by interested parties, 21,000 proposals to combat the slump were sent to the
Reichsbank and the Berlin business school, the majority of which suggested going off gold and
19
adopting expansionary policies.
Consequently, critics of the apologetic view conclude that
energetic political leadership could easily have overcome any lingering residual fears
20
connected with the experience of hyperinflation.
In assessing how probable a continued and significant fear of inflation was in the
context of such a protracted and deep deflation, we also have to consider earlier historical
episodes. How unusual were the early 1930s in the context of German economic history? A
decline in the overall price level was not outside recent experience. Prices of industrial goods
had fallen by almost 20 percent in 1925-6, and the general deflation of the 1880s and early
1890s cannot have faded completely from public memory (table 1). Once a deflation was
underway, it normally persisted for 2 to 5 years. While a full appreciation of the underlying
processes may have been lacking, deflations were not unusual events by the standards of
Germany’s economic history.
T ABLE 1 - DEFLATIONS FROM 1856 T O 1926
length of deflation (years)
CPI (percentage)
price index for investment
goods (percentage)
cumulative net social
product growth
(percentage)
Source: Hoffmann 1965.
19
20
Holtfrerich 1990, p. 70.
Büttner 1989.
1856-59
1862-65
1874-79
1882-87
1892-94
1925-26
3
-15.9
-18.6
3
-5.6
-8.3
5
-11.6
-37.4
5
-7.7
+2.2
2
-7.5
-4.3
2
-0.1
-18.9
+5.8
+8.2
-0.6
+20.1
+6.4
-0.7
6
What is unusual about the years 1929-32 compared to earlier episodes is the magnitude of the
deflation. The consumer price index fell by 22.5 percent, and the price of investment goods
21
was reduced by an average of 48.3 percent. NSP shrank by a cumulative 19.2%, whereas in
four out of six previous deflations since the 1850s, it had continued to grow.
MEASURING INFLATIONARY EXPECTATIONS DURING THE G ERMAN SLUMP
In the absence of survey data or unambiguous testimony in historical sources, a number of
indirect methods can be used to infer inflationary expectations. This section briefly describes
the estimation procedures, and applies them to the problem of inflationary expectations in
Germany during the Great Depression.
A standard way of modelling inflationary expectations was first introduced by Mishkin,
who uses a rational expectations approach to infer inflationary expectations from bond prices.
Mishkin begins with Fisher’s equation – the interest rate on a bond will be equal to the sum of
the ex ante expected rate of inflation and the real return anticipated by investors:
it ≡ rrt + π te
22
(1)
where it is the nominal interest rate earned by holding a bond until time t, rrt is the expected
real return to holding the bond, and π t is the inflation rate anticipated by the bond market
e
over the interval t-1 to t.
Since anticipated and actual rates of inflation can diverge, the interest earned by
investors ex post may be higher or lower than the one expected at t-1. The realized return on a
bond held from t-1 to t is
eprrt ≡ it − π t = rrt + π te − π t
(2)
[eprrt is the actual return, πt is actual inflation between t-1 and t]
The mismatch between actual and expected inflation is captured by π t-πt. Economic agents
e
will attempt to forecast actual inflation given the information that they have available at time
21
Hoffmann 1965, table 148, p. 601. The figure for investment goods refers to the years 1929-33, as Hoffmann
gives no figures for 1931 and 1932.
22
Mishkin 1981.
7
t-1. Under the assumption of rational expectations, they will not consistently and continuously
misjudge actual inflation. Hence, the mean expected difference between actual and
anticipated inflation, conditional on the information set Ω at time t-1, should be zero:
E( π t-πtΩ t-1)=0
e
(3)
Forecast errors of rational economic agents – as captured by bond prices – should be
uncorrelated with information available at t-1. The information available to later historians Xt
is a subset of the full wealth of detail Ω t that historical agents were able to use. If Xt is a valid
predictor of real interest rates, we have
rrt= β X t +ut
(4)
where ut is the error term. Combining equations (4) and (2) yields
eprrt= β X t + ut- (π t-πt) = β X t + ut- εt
e
(5)
Both εt and ut will be orthogonal to the information set Xt; we can therefore estimate equation
(5).
Before we can apply the Mishkin approach to Weimar Germany, we need to establish
that the Fisher equation holds, i.e. that there is evidence to support the expectations
hypothesis. Given that the null of non-stationarity for both interest rates and inflation rates
(measured by the price index of industrial output) cannot be rejected (using both ADF and the
DF tests), the Fisher equation in its weak form requires that interest and inflation rates share a
23
common stochastic trend. The Johansen maximum likelihood procedure, allows us to reject
the null hypothesis of no cointegrating vector against the alternative of one or more vectors
(test statistic 21.58 (monthly rate) and 26.3 (gold bonds) vs. a 95% threshold level of 19.96,
sample 1926/ 1-1931/ 6), while the hypothesis of at most one cointegrating vector could not be
24
rejected. Thus we find that interest rates and inflation cointegrate, a finding that is reinforced
by the fact that the residuals of a regression of interest rates on price changes appear to be I(0)
23
The bond data for 6% issues is from Konjunkturstatistisches Handbuch 1936, p. 119. Gaps in the series were
interpolated using the private discount rate.
8
25
(the Augmented Dickey-Fuller test is significant at the 95% level, ADF = -3.04**). Note that
most of the tests for non-stationarity reported here refer to sample periods that are shorter
than is customary. Recent research strongly suggests that the use of high-frequency samples
reduces this problem – there are significant increases in the power of ADF tests with higher
26
sampling frequency. Nonetheless, our results should be treated cautiously.
Along with other authors, we find that the coefficient on the inflation variable is less
27
than unity (the midpoint estimate is 0.75).
Note, however, that the coefficient is not
significantly different from one – a χ test yields a statistic of 0.35, which is insufficient to reject
2
the restriction at customary levels. The same is true if the cointegrating vector from a VAR
28
with 6 lags is used.
Tax effects are probably responsible for finding coefficients less than
29
unity. Note also that our estimate of the coefficient on the inflation variable is smaller (yet not
statistically significantly so) than coefficients found in the case of three high-inflation countries
30
in the Third World (Argentina 1976-87, Brazil 1971-84, and Mexico 1979-1991).
We follow Cecchetti’s approach to the implementation of the Mishkin method.
31
Included in the OLS regression are the nominal interest rate, as well as the money supply,
output, and prices. The actual specification of a variable x is ∆x= log(xt/ xt-12).
32
Instead of using
the first, thirteenth and twenty-fifth lag, as Cecchetti does, we implement Hendry’s general-to-
24
These results were generated using a maximum lag of 3 months in the VAR. Series have a mean, the
cointegrating equation has an intercept, but no series has a trend.
25
Also, producer prices appear to granger-cause monthly interest rates (one month lag, 6% significance) and the
interest rate on gold bonds (0.1% significance).
26
Choi and Chung 1995, Hooker 1993.
27
Evans and Lewis 1995, Mishkin 1992, Gagnon 1996. Crowder and Hoffmann (1996) find evidence of a unit
coefficient in the US case.
28
Microfit 2.0 was used for VAR estimation.
29
Summers 1983. Note that the finding of a coefficient less than unity implies that the German bond market did
not react with extreme sensitivity to increases in actual inflation before the outbreak of the crisis. This suggests
that the memory of hyperinflation was particularly potent given the political turmoil during the later period,
whereas the relative calm in the late 1920s underpinned a relatively mature bond market. Note that the term
structure of bonds and deposits, however, suggests that even during the 1920s, fear of a resurgence of inflation
persisted. Cf. Balderston 1991, 1992.
30
Cf. Phylaktis and Blake 1993, table 2, p. 595.
31
Cecchetti (1992, 1989). As Cecchetti (1992, p. 148; cf. also Mishkin 1992, appendix 2) points out, the only
other necessary condition is that the real interest rate is stationary. For the period preceding the forecasts, 19261930, we can reject the null of non-stationarity at the 90 percent level using an ADF(3) test. Note, however, that
the sample periods for all tests for non-stationarity reported here are much shorter than is customary. Even if the
use of high-frequency samples remedies this problem somewhat (on the significant increase in the power of ADF
tests with high sampling frequency, cf. Choi and Chung 1995) our results in this regard should be treated
cautiously. On the power of ADF tests in short samples with high frequency, cf. also Hooker 1993.
32
Similar to Cecchetti (1992, p. 149), we do not chose this transformation for the nominal interest variable.
9
specific modelling approach, estimating a full set of a maximum of 13 lags, and then testing
33
down.
If we were to use the fitted values from an OLS regression as the expected real interest
rate level, this would imply that economic agents 'knew' the right specification for the entire
sampling period. This is clearly unrealistic. We therefore use rolling 1-month ahead forecasts
of the real interest rate to infer ex ante rates – i.e. the initial model is estimated for 1926/ 8 to
1930/ 1, and on the basis of this estimate, a forecast for 1930/ 2 is calculated. In the next
round, the sample period for initial estimation is expanded to 1926/ 8 - 1930/ 2, and on the
basis of the new equation, a forecast for 1930/ 3 is generated. Agents are therefore expected to
only 'know' the data that is available to them up to the point in time when they make their
'forecast'. Estimation results for the initial periods are:
T ABLE 2: FORECASTING EQUATIONS - MISHKIN APPROACH
public bonds
(1)
model
Variable
Coefficient Std.Error
i(-1)
π(-1)
π(-4)
m(-1)
m(-3)
m(-8)
UNIT(-3)
Q(-3)
C
Sample
2
R
LM – F
ARCH - F
Normality χ
Reset F
Note:
t-value
0.773
-0.19
0.89
0.914
0.39
-0.21
0.11
0.066
1.72
0.15
-9.6
-5.4
0.087
5.17
13.9
1.74
-1.87
-0.4
1928/ 7-1930/ 12
0.99
2
monthly rates
(2)
0.2
0.39
1.78
Coefficient Std.Error
t-value
0.86
-1.22
0.29
-0.001
0.09
0.03
0.03
0.0006
9.3
-38.1
9.9
-1.5
0.0006
0.0004
1.4
2.3
1.6
1.8
1926/ 5-1930/ 11
0.986
0.169
0.29
4.6
0.35
Estimation method is RLS. The dependent variable is the real interest rate on public bonds (model
1) and on monthly money (model 2).
The two specifications in table 2 use different dependent variables – either the average return
34
on public bonds, or the monthly interest rate from the interbank market. Differences in the
33
Cecchetti 1992, p. 149. Cecchetti (1992, 1989) was able to use a considerably larger dataset, spanning the years
10
sample period used are due to the maximum lag of a dependent variable that was included in
the models. In both equations, no significant coefficient on output could be found, and the
variable was consequently excluded from the regression equations. Both models pass a
number of tests. In particular, they are free from autocorrelation, as evidenced by the low Fstatistics on the LM tests. Heteroscedasticity-constistent standard errors are almost identical to
the common standard errors. The residuals of our regression are normally distributed. Resettests give no evidence of model misspecification. We ultimately favour the model based on
public bonds because the normality of residuals is more firmly established. Also, there is
another reason for preferring the public bond data. As the banking sector faced a liquidity
crunch in the summer of 1931, the fundamental assumption underlying the Mishkin approach
– i.e. that the real interest rate as well as liquidity premia are constant – begins to be
questionable. In the case of interbank rates, it is clear that liquidity is at a premium during the
summer of 1931. Bond yields, however, should be much less affected by short-term liquidity
needs and show little change during the banking crisis.
35
Ultimately, we are not only interested in measuring expected inflation. The purpose of
the exercise is to gauge the room for 'manoeuvre' during the crisis. Hence, we are ultimately
interested in the level of expected inflation conditional on a change in policy that would have
attempted to slow the economic collapse. This is not directly observable. As an alternative, we
can examine periods when such a new course of action was regarded as more likely – such as
Britain leaving the gold standard, the introduction of currency controls, and electoral
successes of the Nazi party. If our estimates of expected inflation increase during these periods
(and diverge significantly from actual price changes), then it is possible to argue that a new
policy might have triggered rapid increases in inflationary expectations. Figure 1 gives the
∧
implied inflation rate from our preferred model (1), where π te = i − epprt (using our preferred
model specification (1) with monthly updating of the information set):
1919 to 1940, and could consequently estimate with a maximum lag of 25 months.
34
Konjunkturstatistisches Handbuch 1936, p. 112. Note that our main findings do not depend on the
interpolated values during the second half of 1931, as innovations were already large and significant before
markets were closed temporarily.
35
The issue of possible changes in liquidity preference is dealt with in more detail below. I thank Peter Temin for
insisting that I examine this point in more detail.
11
10
5
inflation in %p.a.
0
-5
-10
-15
forecast
95% upper bound
95% lower bound
actual
-20
-25
1931-1
1931-7
1932-1
1932-7
1933-1
1933-7
1934-1
1934-7
1935-1
1935-7
FIGURE 1 - O NE-STEP-AHEAD FORECAST OF EXPECTED INFLATION AND ACTUAL INFLATION (FROM MODEL 1,
TABLE 1)
From the summer of 1931 onwards, expectations become highly uncertain. While the mean
forecast is only marginally above actual inflation, the confidence intervals widen significantly.
At a time when prices were falling at about 10 % (annualized), the Mishkin model suggests
that (i) expectations of as little as 0.5 % deflation p.a. would have been compatible with basic
rationality during some months in 1931/ 1932, and (ii) after a long period when inflation
forecasts were anticipated correctly to within a few percentage points, a very wide range of
outcomes was possible.
Germans were indeed jittery about a resurgence of inflation. Figure 2 plots the
standard error of the inflation forecast. The sharp rise in uncertainty begins with the banking
crisis, accelerates with the British departure from gold, and is only reversed once the
reparations settlement in Lausanne has been achieved. As soon as exogenous policy events
such as the banking crisis or the British decision to abandon the gold standard increase the
likelihood of the German government adopting a more active policy, inflationary expectations
appear to have increased.
12
5
1
2
3
4
4.5
4
s.d. of forecast
3.5
(1)
(2)
(3)
(4)
3
Banking crisis
England abandons gold
Lausanne conference
Nazis seize power
2.5
2
1.5
1
0.5
0
19
31
-1
19
31
-4
19
31
-7
3
19
1-
10
19
32
-1
19
32
-4
19
32
-7
3
19
2-
10
19
33
-1
19
33
-4
19
33
-7
3
19
3-
10
19
34
-1
19
34
-4
19
34
-7
3
19
4-
10
19
35
-1
19
35
-4
19
35
-7
FIGURE 2 – STANDARD DEVIATION OF INFLATION FORECAST (FROM MODEL 1, TABLE 1)
Note that, for an important part of this period, data on bond prices and yields is not directly
available. Stock exchanges were closed in the fall of 1931 and early 1932. We used the market
prices established in telephone trading, as recorded by the Reichsbank.
36
However, since the
disturbances are clearly visible long after the interpolation period, it is evident that our results
are not driven by this. Also, forecasts from a model using only monthly interest rates – which
are available without interruption – give broadly similar results.
Yield Spreads
It could be argued that the conditions of the German bond market following the banking crisis
were highly unusual, and that the changes in yields did not reflect a fear of increased inflation
in the future. Rather, it measures a change in liquidity premia, differences in default risk,
movements in the real rate of interest, or time-varying risk premia. We can resolve the issue of
liquidity risk by examining the spread of mortgage bonds (Pfandbriefe) and rye bonds over
public bonds. Not all bonds will be affected equally by changes in liquidity premia. Pfandbriefe,
for example, are notoriously illiquid because of small issue size and limited free float. The
36
Beer 1998.
13
same is not true of public bonds, which were predominantly issued in large tranches by the
Reich or individual Länder. Since the possibility of an internal default of the Reich was very
small even in 1931, and Pfandbriefe had a history of zero default since their inception, the
spread between the two will largely reflect differences in two factors, i.e. liquidity premia and
37
the 'gold clause'.
The bond market was devastated by the effects of the hyperinflation. A strong and
stable demand for bonds disappeared together with the German rentier (Balderston 1993). All
issuers found it difficult to tap the market for longer maturities or for larger issues. These facts
alone indirectly suggest that strong fear of resurgent inflation was never far from the surface.
This view is reinforced by the fact that a very large number of bonds – virtually all except
issues by the state – had 'gold clauses', which guaranteed the value of coupon and principal
payments in the case of another departure from the gold standard. The vast majority of bonds
issued during the Weimar period thus contained a hedge against inflation, albeit an imperfect
and indirect one – gold clauses could always be abrogated by parliament.
During the period 1919-23, bonds denominated in sugar, rye, coal, gold and
potassium had been issued. By the late 1920s, most of these bonds had been redeemed. Rye
38
bonds, however, continued to be traded. They also make an ideal benchmark against which
inflationary expectations can be judged – the stabilization of the Mark in 1923/ 24 was carried
out using rye to back the currency. Bonds denominated in rye were therefore regarded as the
ultimate hedge against inflation.
We can use the difference in the yields of these instruments to examine inflationary
expectations. Bonds with a 'gold clause' can be thought of as containing a conditional put
option on repayment in gold. Thus, any increase in inflation should make these instruments
relatively more valuable than bonds without the option – even if the devaluation has not yet
occurred, we can think of the rise in ex ante inflation as increasing the probability that the
value of the underlying will be below the strike value of the option. While prices are rising, the
yield of the gold-backed bonds should be lower than that of the instrument which merely
guarantees nominal repayment.
39
If rye prices are flat, then the yield spread between Mark
and rye-denominated bonds will ceteris paribus be an indicator of anticipated inflation.
37
An internal default was never considered a possibility, even by the most pessimistic observers. In contrast, an
external default was certainly much more likely once the crisis deepened.
38
Wolfgang 1931, p. 131-2.
14
3.50
3.00
2.50
In %
2.00
1.50
1.00
0.50
0.00
-0.50
1928
1929
1930
1931
1932
1933
1934
1935
time
FIGURE 3 - SPREAD OF GOVERNMENT BONDS OVER GOLD-BACKED BONDS
There are thus two factors working in opposite directions - after July 1931, the change in
liquidity needs ought to have increased the spread of Pfandbriefe and rye bonds over
Reichsanleihen (or diminished a negative spread), whereas increasing fear of inflation should
have caused a rise in the difference. Figure 3 shows that the differential widens dramatically,
40
with the more liquid Reichsanleihen carrying higher yields to maturity. Hence the (positive)
spread of Reichsanleihen over Pfandbriefe is a lower bound on the change in inflationary
expectations, as this is the net spread, i.e. after liquidity effects have presumably tightened the
spread from the level it would have reached if Pfandbriefe had been equally liquid.
41
Rye
bonds show a broadly similar pattern during the crucial period 1931/ 32. In July 1931, the
39
Note that the rise in French interest rates after 1931 is also driven by expectations of devaluation (Hautcoeur
and Sicsic 1998, p. 18).
40
Since the German exchanges were closed from August 31 to April 32 (having briefly reopened in September),
we had to use other sources for yields during this period. Securities continued to be traded by telephone or in
informal gatherings; the Frankfurter Zeitung published selected prices for the period of the stock exchange closure in
its issue from 12th April 1932. We use the price of the 1929 Hilferding loan, which was relatively frequently
traded and liquid, to interpolate the price series for public bonds. Montly prices of the Hilferding loan used for
the interpolation for the period 1929-31 were collected from Frankfurter Zeitung. For August, no bond prices were
published; instead, interpolation with the monthly money rate was carried out.
41
Note that, because the data is only available in aggregate form, it is not possible to match maturities perfectly.
However, since the number bonds from which the index is derived is large, it is unlikely that dramatic changes in
sample composition over short periods of time determined the results.
15
yield spread of public bonds over gold bonds was 0.75%, the spread over rye bonds 0.82%. By
September 1931, when the trading floors were reopened for a brief period, the differential had
42
widened to 0.83% and 1.98%, respectively.
In April 1932, when continuous trading
resumed, the spread jumped to 2.75% and 3.49%, after which it began to decline gradually.
The evidence from yield spreads between assets that guarantee nominal and real returns is
therefore consistent with our finding that fear of inflation cannot be ruled out given the other
evidence from the bond market, and that it was strongest in the second half of 1931 and the
first half of 1932.
GARCH models
Investors observe price changes at a rate π at time t. Also, they have access to the information
set up until time t-1. Based on these observations, economic agents will be able to form a
prediction of the inflation rate π conditional on Ω as well as a conditional variance forecast h.
The difference between the expected and realized inflation rate will be ε = π- π . How does the
e
conditional variance of the inflation forecast vary with the arrival of 'news' about past inflation
and other factors? Following the work of Engle and Bollerslev, modelling the uncertainty of
economic variables through autoregressive conditional heteroskedasticity models (GARCH)
has become increasingly common. The generalized form is
p
q
i =1
i =1
ht = ω + ∑ α i ε t2−1 + ∑ β i ht −1
(6)
where ω, α and β are parameters. In a GARCH model with p= q=1, the effect of a single
shock dies out gradually over time. The adjustment parameter α measures the speed with
which the conditional variance of the inflation forecast changes as a result of an unexpected
change in inflation, and β measures the extent to which a shock to conditional volatility
persists.
43
In empirical work, GARCH (1,1) models have been preferred in most cases. Table 3
shows the results for estimating Garch models for two sample periods: January 1925 to
September 1935, and July 1931 to April 1932. Because price changes violate the normality
42
Note that the price of Goldpfandbriefe and of rye bonds did not remain constant – the value of the data is not
undermined by non-trading. Data on rye bonds is from Statistisches Reichsamt (various years).
43
Bollerslev 1992. To test for the possibility of asymmetric responses, EGARCH and TGARCH estimation was
carried out, but the leverage factor turned out to be insignificant.
16
assumption, we used the robust covariance matrix procedure by Tim Bollerslev and Jeff
44
Wooldridge.
T ABLE 3
ESTIMATES OF C ONDITIONAL INFLATIONARY UNCERTAINTY: GARCH (1,1)-MODELS
regression
sample
period
ω
α
β
1
2
3
4
1926:01-1935:09
1926:01-1935:09
1931:07-1932:04
1931:07-1932:04
9.9
(2.15)***
1.42
(0.065)***
-0.69
(0.032)***
0.28
(0.12)**
1.17
(0.139)***
-0.066
(0.099)
0.078
(0.077)
3.11
(1.59)***
-0.024
(0.04)
0.12
(0.055)
2.89
(1.56)*
-0.11
(0.04)
Likelihood
351
343
14.6
11.7
Akaike
6.1
5.96
3.72
3.34
*
indicates significance at the 90% level
**
indicates significance at the 95% level
**
indicates significance at the 99% level
Notes:
Standard errors in parentheses. Estimation method is maximum likelihood with Bollerslev-Wooldridge
robust standard errors and covariances. Equations (2) and (4) use a dummy variable (=1 for January
1932, 0 otherwise) to capture the effect of the fourth emergency degree
The coefficient α is larger than unity throughout, suggesting that unexpected increases in
inflation lead to unusually large upward revisions of the one-step-ahead forecast of the
conditional variance. Weimar investors obviously grew extremely nervous once actual
inflation began to depart from their inflation forecast. Equations 2 and 4 use a dummy for the
fourth emergency decree, which reduced the price level by governmental fiat. Persistence of
shocks, measured by β, varies between samples, but is never large once we control for the
effect of government intervention in the price setting process.
Regressions 3 and 4 report results for the period starting with the banking crisis until
the reopening of the exchanges in April 1932, during which Britain left the gold standard and
the most severe austerity measures (such as the fourth emergency decree) were implemented.
While the results for the shorter period have to be treated with care, the findings are
nonetheless striking. The coefficient on α is markedly larger than before, indicating that
conditional variances increase much more sharply in response to a larger mismatch between
predicted and actual rates of inflation than during the relatively calm period until the summer
of 1931.
44
Bollerslev and Wooldridge 1992.
17
T he T erm Structure of Interest Rates
The term structure of interest rates is known to contain information about future changes in
45
inflation.
A standard way to examine the extent to which changes in inflation between
periods m and n can be predicted from the slope of the term structure is to estimate:
[
]
π mt − π nt = α m,n + β m,n itm − itn + η mt ,n
(7)
where π denotes the rate of inflation, i refers to the interest rate, α is the intercept, and η is the
residual. αm,n will reflect the difference between ex ante real rates of return between the two
46
periods. As information on the exact maturity of long-term bonds is unavailable, the analysis
has to be adapted to the data constraints. For imperial and interwar Germany, we can use the
47
long-term bond yields as compiled by the Bundesbank.
The private discount rate is also
available for pre-war and interwar period. The results of applying equation are summarized
in table 4.
T ABLE 4: TERM STRUCTURE EQUATIONS
model
(1)
Variable
(2)
Coefficient
Std.Error
t-value
Coefficient
Std.Error
t-value
∆π-1
∆i-1
-0.47
1.22
0.159
0.64
-2.94
1.91
-0.44
1.22
0.147
0.629
-3.03
1.93
C
-0.54
0.6
-0.89
-0.699
0.584
-1.1
Sample
2
R
Heterosc.-F
LM - F
ARCH - F
Normality χ
Reset F
2
Note:
45
1880-1913
0.25
1.76
2.1
0.19
1.53
1.08
1880-1929
0.25
1.86
1.58
0.23
1.1
1.13
Estimation technique is OLS, which, under plausible assumptions (Mishkin 1990), will yield
unbiased and efficient estimates.
Mishkin 1990b, c. The abundant literature on the topic cannot be reviewed in this paper. Classic papers also
include Fama (1990) and Campbell and Shiller (1987).
46
Mishkin 1990c, p. 78-81.
47
Bundesbank 1976, p. 278.
18
For both sample periods used in table 4, the slope of the term structure has predictive power.
It is significant at the 6 % level. Both models pass a number of tests. They are free from
autocorrelation, show no evidence of ARCH or heteroscedasticity, and Ramsey's reset test (up
to the second power) does not suggest any misspecification.. We conclude that the difference
between bond yields and the private discount rates forecasts changes in inflation. If our earlier
argument is true, and Germans were afraid of inflation during the Great Depression, then this
should be reflected in a particularly steep yield curve. The long-term mean of the difference
between bond yields and the private discount rate is 0.97. In 1931, it jumps to 3.45% (from
0.22% in 1930), and to 3.32% in 1932. Based on model (1) in table 4, this implies increases in
inflation by 2.52% and 5.3% p.a. Information from the yield curve therefore reinforces our
earlier finding that Germans expected accelerating price increases in 1931/ 32.
Futures prices
Further corroborating evidence comes from the price of futures contracts for foreign exchange
and agricultural commodities. Let Sj,t be the spot rate of good j at time t, and Fj,t the price of a
one-period ahead forward contract. With investors neutral to risk and efficient markets,
[
Fj ,t = Et S j ,t +1
]
(8)
48
where Et denotes expectations. Inflationary expectations can then be measured as
π et =
12
( f j ,t − s j ,t + s )
T
(9)
where lower case letters denote natural logarithms, and T is the number of months to
49
settlement.
We can use this method for both forward Reichsmarks and the price of two
50
agricultural commodities, wheat and rye.
A discount on forward Marks implies that agents
51
During the years 1919-23, a decline in the Mark's
expect the exchange rate to depreciate.
external value was often viewed as synonymous with an increase in the rate of inflation. Since
Frenkel's seminal contribution, inflationary expectations during the German hyperinflation
48
We are abstracting from cost-of-carry.
Hamilton 1992.
50
A sceptical view -- at a very general level -- can be found in Mishkin (1990a). The results in our own work, as
well as in the literature (Hamilton 1992) suggest that the problems he identifies are less severe than might be
supposed.
51
That is, if expectations are rational in the sense of Muth (1960).
49
19
52
have routinely been modelled using forward rates.
For the period from February 1921 to
August 1923, Frenkel demonstrated that β in a regression St = α + βFt −1 is not significantly
53
different from unity.
Germans during the early 1930s would have immediately associated
54
any fall in the external value of the Mark with an imminent rise in inflation. In the German
case, we do not need to subscribe to a theory of imported price inflation. In a country where,
at the height of the hyperinflation, prices had been changed daily even at the retail level based
on the current dollar/ mark exchange rate any sudden fall in the gold value of the mark would
lead to a rapid rise in expected inflation and the velocity of money.
55
After the banking crisis, and immediately before Britain announced that it was going
to leave the gold standard, three-month forward Reichsmarks traded at a discount of more
than 1.6%, implying an annual depreciation of the exchange rate of 6.8%.
56
During the
autumn of 1931, when the markets expected the Reichsmark to follow Sterling's devaluation,
the forward discount widened to as much as 75% annualized – a level not seen since the
darkest days of the hyperinflation, in April 1923, when three month forward Marks stood at
57
an (annualized) discount of 65% (cf. figure 4). Given discounts of this order of magnitude, is
reasonable to suppose that such differences are likely to be dominated by expectations of
58
exchange rate and price changes.
52
Frenkel 1977, Webb 1986, 1989.
Frenkel 1977, p. 655.
54
In the case of the 1926 stabilization of the Franc Poincaré there is debate if there is granger-causality running
from the exchange rate to inflation or vice versa. Sicsic 1982, Eichengreen 1982b.
55
Eichengreen 1982a, p. 135.
56
Most other currencies show rising premia against Sterling, especially in August and September. Cf. Einzig
1937, p. 470-71.
57
Einzig 1937, p. 294. The market rates on forward Reichsmarks are from Einzig 1937, p. 450-70. Note that
there are no observations for August 1931 - the dotted line marks the interpolation from July to September. My
figures for the hyperinflation differ slightly from those in Webb (1986, p. 776-77) since he used one-month
forward rates.
58
Frenkel 1977, p. 654. The bank rate differential vis-a-vis Britain was +5.5% in July and August of 1931; the
difference in call money rates varied between +3.87% in November and +4.3% in June. Given the rate
differential, forward Reichsmarks should have traded at a premium.
53
20
5
0
discount/premium (pf)
-5
-10
-15
-20
-25
-30
-35
-40
Nov.24
May.25
Nov.25
May.26
Nov.26
May.27
Nov.27
May.28
Nov.28
May.29
Nov.29
May.30
Nov.30
May.31
FIGURE 4 - 3 MONTHS FORWARD DISCOUNT / PREMIUM ON R EICHSMARKS, 1924-31
Data on wheat futures also indicate that agents expected price increases, conditional on a
59
major economic policy change. During 1931 and 1932, the future price of rye and wheat as
60
quoted on the Berlin exchange was almost always above the spot price. Instead of comparing
futures and spot prices, which may diverge for a number of reasons such as storage cost etc.,
we follow the literature in comparing futures prices with different settlement dates.
61
The
actual change in spot prices between O ctober and December 1931 was -2.78% (annualized).
In the first week of August, when the December contract began to trade, the expected price
change for the futures contract between O ctober and December is +0.05% (annualized). After
the suspension of the gold standard by the Bank of England, in the last week of August, this
figure had risen to 8.5% p.a. Also, the rate of price increases was expected to accelerate
rapidly. For most of August, the O ctober future was almost always trading below the
September contract, implying that prices were expected to fall over the next month. At
horizons of up to three months, however, there was (unwarranted) fear of inflation at precisely
those times when a change in policy became more likely.
59
Data on rye futures are less useful. While the spot price of wheat has a correlation coefficient of 0.51 with the
price index used above, the figure for rye is 0.14 (1930/ 10-1932/ 11).
60
The data on futures prices are from Statistisches Reichsamt (various years).
21
Summary
Having examined evidence from yield spreads, the Mishkin approach, and futures prices, we
consistently find evidence that fear of inflation can be detected from the summer of 1931
onwards. Inflationary expectations appear highly sensitive to changes in the likelihood of a
more expansionist policy. Given that prices were falling at the time, it may appear surprising
that an approach based on rational expectations (such as the Mishkin model) can generate
sustained departures from actual inflation. The apparent contradiction is actually not
uncommon, and is a variation of the so-called 'peso problem' (forward rates of the peso for
extended periods appear to underpredict the dollar exchange rate). The solution to the puzzle
in the Mexican case – a non-zero probability of a major discontinuity occurring – is actually
also responsible for the German data. Having experienced one hyperinflation within living
memory, economic agents were highly sensitive about possible changes to the monetary
regime or the exchange rate.
Such 'long memory'' is not unusual by the standards of today's OECD countries.
Gagnon examines the impact of past inflation on ten-year bond prices.
62
He finds clear
statistical evidence that expectations about future inflation are strongly influenced by the
experience of inflation over the ten preceding years. Surely, if the relatively mild bout of price
increases in most OECD countries lingered in public memory, the traumatic consequences of
the German hyperinflation should have had a much more lasting impact.
The timing of the upward turn thus reinforces contemporary observations about the
events preceding the banking crisis and Britain’s departure from gold. During a conference
held by the Friedrich List society in September 1931, memories of the summer's events were
still fresh. When exploring opportunities for monetary expansion, one participant commented
how the declining reserves of the Reichsbank in June and July had spurred fear of inflation
63
and aggravated the run on deposits. As the financial adviser at the British Embassy in Berlin
put it: "The reasons against Germany’s departure from the gold standard are sufficiently well
known... The fear of inflation is such that a fall in the quotation of the mark as expressed in
64
dollars would immediately lead to panic...".
61
62
63
64
After the banking crisis, during the few days
Hamilton 1992.
Gagnon 1996, p. 9.
Reg. Rat a.D. Dr. Stern, cit. acc. to Borchardt 1991, p. 144.
Cit. acc. to Borchardt 1985, p. 242.
22
when the stockmarket was open, there was a buying frenzy – driven, as the contemporary
journalists commented, by investors who were trying to buy inflation-proof assets.
65
D ETERMINANTS OF INFLATIONARY EXPECTATIONS AND U NCERTAINTY
Ferderer and Zalewski have recently shown in a cross-section of countries how higher interestrate uncertainty aggravated the Great Depression. We suggest that fear of inflation and
uncertainty about the future price changes also had a direct effect on the range of policy
options available to governments. Cutting the link with gold was particularly hard for the
countries that had experienced monetary turmoil after WWI, and which had spent so many
political and social resources trying to return to gold.
66
Also, the British decision to leave the gold standard was not the 'window of
opportunity' for Germany to follow suit, as some authors have argued. Instead of making this
step easier, Germany's position as a second mover compounded the difficulties. If economic
agents were already nervous about domestic inflation when a foreign country left the gold
standard, it is not difficult to guess how much more rampant 'inflation hysteria' would have
been had the Brüning government (or some other coalition) adopted such a policy. Next, we
need to examine if there was a range of policy options that could have been pursued without
igniting widespread panic. To this end, we estimate multi-factor GARCH model of inflation
uncertainty.
65
66
Bauer 1931, p. 246.
Feinstein, Temin and Toniolo (1997), p.114-20.
23
T ABLE 5
ESTIMATES OF C ONDITIONAL INFLATIONARY UNCERTAINTY: GARCH (1,1)-MODELS
regression
sample period
ω
α
β
revenue
1
2
3
4
1926:01-1935:09
29.3
(5.6)***
0.83
(0.07)***
-0.13
(0.02)***
-0.022
(0.004)***
1926:01-1935:09
33.97
(15.1)**
1.51
(1.67)
-0.75
(0.36)**
-0.02
(0.011)*
1926:01-1935:09
25.24
(5.58)***
1.08
(0.23)***
-0.33
(0.24)
-0.0052
(0.0018)***
-2.46
(0.744)***
1926:01-1935:09
34.5
(16.6)**
1.04
(0.88)
-0.46
(0.69)
discount rate
emergency decree
dummy
expenditure
0.03
(236.9)
0.02
(118)
-0.02
(0.015)
340
6.2
Likelihood
339.4
343.8
332.9
Akaike
6.2
6.3
6.1
*
indicates significance at the 90% level
**
indicates significance at the 95% level
**
indicates significance at the 99% level
Notes:
Standard errors in parentheses. Estimation method is maximum likelihood with Bollerslev-Wooldridge
robust standard errors and covariances.
Table 5 shows the effect of a number of policy instruments. The emergency decree has a small
and statistically insignificant effect on the conditional variance of inflation, and expenditure of
the Reich appears to be insignificant (and wrongly signed) as well. In contrast, Reich revenues
do reduce uncertainty about the course of inflation, and so do higher interest rates. The policy
of trying to balance the budget by cutting expenditures and raising taxes, continued after the
introduction of exchange controls in the summer of 1931, has long been regarded as a
textbook example of bad economic policy. It appears that not all of the Brüning government's
austerity measures were equally harmful. In a country where memories of the hyperinflation
were still vivid, showing fiscal rectitude did apparently calm the nerves of the public. Also,
higher discount rates lead to a reduction in conditional volatility. Hence, the same policy mix
that allowed the stabilization of the currency in 1924 – higher taxes and tight monetary policy
67
– were also useful in reducing inflationary uncertainty.
That this was necessary at all is
largely due to exogenous shocks – the failure to obtain a loan in the summer of 1931 that
could have saved the German banks, and the British departure from gold.
Hence, the debate about policy options during the German slump has been correct in
emphasizing the crucial role of inflationary expectations. Uncertainty and fear of inflation are
clearly visible in all asset markets, and it was through the deflationary – and in all likelihood
67
Sargent 1986.
24
economically harmful – austerity policies of the Brüning government that a minimum degree
of public confidence could be restored. Given the adverse effects of uncertainty in its own
right, there were also direct positive effects as a result of the tax hikes and high interest rates.
Brüning's own comment that he fell 50 metres before the finishing line has often been cited as
a sign of ineptitude – focussing on the issue of reparations instead of the economic collapse.
The sharp reduction in uncertainty on most counts from the summer of 1932 onwards
demonstrates that Brüning's emphasis on the reparations issue was not quite as misguided as
some scholars have argued. Once the Lausanne conference effectively freed Germany from
the shackles of reparations, public confidence in the currency returned. The expansionary
measures adopted in 1932 and 1933 were less likely to fan the flames of inflationary
expectations precisely because the economic and political situation appeared to be stabilizing.
This paper therefore strongly suggests that, far from dealing a deadly blow to the German
economy without a good reason, the Brüning government's economic policy was not only
without an alternative, but may have been beneficial in calming the nerves of the public at a
critical juncture, and preparing the ground for the expansion that started in the second half of
1932.
25
References
Balderston, Theodore (1991), 'German Banking Between the Wars: The Crisis of the Credit Banks', Business
History Review 65, 554-605.
___ (1993), The Origins and Course of the German Economic Crisis, Haude & Spener: Berlin.
Barkai, Avraham (1990), Nazi Economics. Ideology, Theory and Policy, New Haven: Yale University Press.
Bauer, Lothar (1931), 'Die Börsen', Die Wirtschaftskurve 10/ III, 246-51.
Beer, Joachim (1998), Der Funktionswandel der deutschen Wertpapierbörsen in der Zwischenkriegszeit (19241939), Peter Lang: Frankfurt.
Bernanke, Ben and Mark Gertler (1990), 'Financial Fragility and Economic Performance', Quarterly Journal of
Economics 105, pp. 87-114.
___ (1989), 'Agency Costs, Net Worth, and Business Fluctuations', American Economic Review 79, pp. 14-31.
Bittlingmayer, George (1998), 'Output, Stock Volatility and Political Uncertainty in a Natural Experiment:
Germany, 1880-1940', Journal of Finance.
Black, Fischer (1997), 'Interest Rates as Options', Journal of Finance 50, pp. 1371-76.
Bollerslev, Tim, and Jeffrey M. Wooldridge (1992), 'Quasi-Maximum Likelihood Estimation and Inference in
Dynmaic Modles with Time-Varying Covariances', Economic Reviews 11.
Bollerslev, Tim, Ray Chou, and K. Kroner (1992), 'ARCH Modeling in Finance: A Review of the Theory and
Empirical Evidence', Journal of Econometrics 52.
Bombach, Gottfried (1976), 'Einleitung', in Gottfried Bombach, ed., Der Keynesianismus II: Die
beschäftigungspolitische Diskussion vor Keynes in Deutschland. Dokumente und Kommentare, Berlin:
Springer.
Borchardt, Knut (1991), Perspectives on Modern German Economic History and Policy, Cambridge:
Cambridge University Press.
___ (1985), 'Das Gewicht der Inflationsangst in den wirtschaftspolitischen Entscheidungsprozessen während der
Weltwirtschaftskrise', in: Gerald D. Feldman, ed., Die Nachwirkungen der Inflation auf die deutsche
Geschichte, Munich: Oldenbourg.
___ and Hans-Otto Schötz (1991), eds., Wirtschaftspolitik in der Krise. Die (Geheim-) Konferenz der FriedrichList-Gesellschaft im September 1931 Möglichkeiten und Folgen einer Kreditausweitung, Baden-Baden:
Nomos.
___ and Albrecht Ritschl (1992), 'Could Brüning Have Done It? A Keynesian Model of Interwar Germany,
1925-1938', European Economic Review 36, pp. 695-701.
Brayton, Flint, Andrew Levin, Ralph Tryon, and John Williams (1997), 'The Evolution of Macro Models at the
Federal Reserve Board', Carnegie-Rochester Conference Series on Public Policy.
Brockhaus Enzyklopädie (1929), Vol. 4, Artikel 'Deflation', Brockhaus Verlag: Leipzig.
Bresciani-Turroni, Costantino (1937), The Economics of Inflation. A Study of Currency Depreciation in PostWar Germany, London.
Brüning, H. (1970), Memoiren (Stuttgart).
Büttner, U. (1989), 'Politische Alternativen zum Brüningschen Deflationskurs', in: Vierteljahrshefte für
Zeitgeschichte 37.
Campbell, John and Robert Shiller (1987), 'Cointegration and Tests of Present Value Models', Journal of
Political Economy 95, pp. 1062-1088.
Cecchetti, Steven (1992), 'Prices During the Great Depression: Was the Deflation of 1930-1932 Really
Unanticipated?', American Economic Review 82.
___ (1988), 'Deflation and the Great Depression', unpublished manuscript, University of Ohio.
Choi, In and Bhum Chung (1995), 'Sampling Frequency and the Power of Tests for a Unit Root: A Simulation
Study', Economic Letters 49, pp. 131-36.
Crowder, William and Dennis Hoffman (1996), 'The Long-Run Relationship between Nominal Interest Rates
and Inflation: The Fisher Equation Revisited', Journal of Money, Credit and Banking 28, pp. 102-118.
Deutsche Bundesbank (1976), Deutsches Geld- und Bankwesen in Zahlen 1876-1975, Frankfurt am Main:
Knapp.
26
Dominguez, Kathryn, Ray Fair and Matthew Shapiro (1988), 'Forecasting the Depression: Harvard versus Yale',
American Economic Review 78, pp. 595-612.
Eichengreen, Barry (1992a), Golden Fetters. The Gold Standard and the Great Depression, Oxford: Oxford
University Press.
___ (1992b), 'Comment: More Speculation about Destabilizing Speculation', Explorations in Economic History
29, pp. 93-8.
___ and Jeffrey Sachs (1985), 'Exchange Rates and Economic Recovery in the 1930s', Journal of Economic
History 45, pp. 925-946.
Einzig, Paul (1937), The Theory of Forward Exchange, London: Macmillan.
Engle, Robert and Victor Ng (1993), 'Measuring and Testing the Impact of News on Volatility', Journal of
Finance 48.
Evans, Martin and Karen Lewis (1995), 'Do Expected Shifts in Inflation Affect Estimates of the Long-Run Fisher
Relation?', Journal of Finance 50, pp. 225-53.
Fama, Eugene (1990), 'Term Structure Forecasts of Interest Rates, Inflation and Real Returns', Journal of
Monetary Economics 25, pp. 59-76.
Ferderer, J. Peter and David A. Zalewski (1994), 'Uncertainty as a Propagating Force in the Great Depression',
Journal of Economic History 54, pp. 825-49.
___ (1999), 'To Raise the Golden Anchor? Financial Crises and Uncertainty During the Great Depression',
Journal of Economic History 59, pp. 624-58.
Feinstein, Charles H., Peter Temin and Gianni Toniolo (1997), The European Economy Between the Wars,
Oxford: Oxford University Press.
Foreman-Peck, James, Andrew Hughes Hallet and Yue Ma (1996), 'Optimum International Policies for the
World Depression 1929-1933', Economies et Societé 22, pp. 219-42.
Frenkel, Jakob A. (1977), 'The Forward Exchange Rate, Expectations, and the Demand for Money: The
German Hyperinflation', American Economic Review 67, pp. 653-671.
Gagnon, Joseph (1996), 'Long Memory in Inflation Expectations: Evidence from International Financial
Markets', Board of Governors of the Federal Reserve System. International Finance Discussion Papers
538.
Gielen, Gregor (1994), Können Aktienkurse noch steigen? Wiesbaden: Gabler.
Hamilton, James D. (1987), 'Monetary Factors in the Great Depression', Journal of Monetary Economics, 19, pp.
145-169.
___ (1992), 'Was the Deflation During the Great Depression Anticipated? Evidence from the Commodity
Futures Market', American Economic Review 82, pp. 157-177.
Hautcoeur, Pierre-Cyrille and Pierre Sicsic (1998), 'Threat of a Capital Levy, Expected Devaluation and Interest
Rates in the France During the Interwar Period', Banque de France - Notes d'études et de recherche 50.
Hendry, David F. (1993), Econometrics. Alchemy or Science? Essays in Econometric Methodology, Oxford:
Blackwell.
Hoffmann, Walther G. (1965), Das Wachstum der deutschen Wirtschaft seit der Mitte des 19. Jahrhunderts,
Berlin/ Heidelberg: Springer.
Holtfrerich, Carl-Ludwig (1990), 'Economic Policy Options and the End of the Weimar Republic', in: Ian
Kershaw, ed., Weimar: Why Did German Democracy Fail?, New York: St. Martins.
Hooker, Mark (1992), 'Testing for Cointegration: Power versus Frequency of Observation', Economic Letters 41,
pp. 359-62.
Institut für Konjunkturforschung (1935), Konjunkturstatistisches Handbuch, Hamburg: Hanseatische
Verlagsanstalt.
James, Harold (1986), The German Slump, Oxford: Oxford University Press.
Kroll, Gerhard (1959), Von der Weltwirtschaftskrise zur Staatskonjunktur, Berlin: Dunker & Humblot.
Krugman, Paul (1998), 'Japan's Trap', unpublished manuscript, Massachusetts Institute of Techonology.
Mishkin, Frederic (1981), 'The Real Interest Rate: An Empirical Investigation', Carnegie-Rochester Conference
Series on Public Policy 15, pp. 151-200.
___ (1990a), 'Can Futures Market Data Be Used to Understand the Behavior of Real Interest Rates?', Journal of
Finance, pp. 245-57.
27
___ (1990b), 'The Information in the Longer Maturity Term Structure about Future Inflation', Quarterly
Journal of Economics 105, pp. 815-28.
___ (1990c), 'What Does the Term Structure Tell Us About Future Inflation?', Journal of Monetary Economics
25 (1990), pp. 77-95.
___ (1992), 'Is the Fisher Effect for Real? A Reexamination of the Relationship between Inflation and Interest
Rates', Journal of Monetary Economics 30, pp. 195-215.
Muth, J. (1960), 'Optimal Properties of Exponentially Weighted Forecasts', Journal of the American Statistical
Association 55, pp. 299-306.
Phylaktis, Kate and David Blake (1993), 'The Fisher Hypothesis: Evidence from Three High Inflation
Economies', Weltwirtschaftliches Archiv 29, pp. 591-99.
Ritschl, Albrecht (1997), Deutschlands Krise und Konjunktur 1924-1934. unpublished manuscript, Economics
Department, Munich University.
Robbins, Lionel (1937), 'Foreword', in: Costantino Bresciani-Turroni, The Economics of Inflation. A Study of
Currency Deprecation in Post-War Germany, London.
Romer, Christina (1992), 'What Ended the Great Depression?', Journal of Economic History 52, pp. 757-84.
___ and David Romer (1994), 'What Ends Recessions?', NBER Macroeconomics Annual, pp. 13-57.
Sargent, Thomas (1986), 'The End of Four Big Inflations', in: Thomas Sargent, ed., Rational Expectations and
Inflation, New York: Harper and Row, pp. 40-109.
Schwert, G. William (1989), 'Tests for Unit Roots: A Monte Carlo Investigation', Journal of Business and
Economic Statistics 7.
Schwarz, Gideon (1978), 'Estimating the Dimension of a Model', Annals of Mathematical Statistics 6, pp. 461-4.
Sicsic, Pierre, 'Was the Franc Poincaré Deliberately Undervalued?', Explorations in Economic History 29, pp.
69-92.
Statistisches Reichsamt (various years), Wirtschaft und Statistik, Berlin.
Summers, Lawrence (1983), 'The Non-Adjustment of Nominal Interest Rates: A Study of the Fisher Effect', in:
James Tobin, ed., Symposium in Memory of Arthur Okun, Washington: Brookings Institute.
Taylor, John B. (1998), 'An Historical Analysis of Monetary Policy Rules', unpublished manuscript, Economics
Department, Stanford University.
___ (1993), 'Discretion Versus Policy Rules in Practice', Carnegie-Rochester Conference Series in Public Policy
39, pp. 195-214.
Temin, Peter (1976), Did Monetary Forces Cause the Great Depression?, New York: Norton.
___ (1990), Lessons from the Great Depression, Cambridge, MA: MIT Press.
___ (1971), 'The Beginning of the Great Depression in Germany', Economic History Review 24, pp. 240-248.
___ and Barrie Wigmore (1990), 'The End of One Big Deflation', Explorations in Economic History 27, pp. 483502.
Voth, Hans-Joachim (1993), 'Wages, Investment and the Fate of the Weimar Republic: A Long-Term
Perspective', German History 11, pp. 265-292.
___ (1994), 'Zinsen, Investionen und das Ende der Großen Depression in Deutschland', Zeitschrift für
Wirtschafts- und Sozialwissenschaften 114, pp. 267-81.
___ (1995), 'Did High Wages or High Interest Rates Bring Down the Weimar Republic?', Journal of Economic
History 55 (4), pp. 801-21.
Webb, Steven (1989), Hyperinflation and Stabilization in Weimar Germany, Oxford: Oxford University Press.
___ (1986), 'Fiscal News and Inflationary Expectations in Germany after World War I', Journal of Economic
History 46, pp. 769-794.
Wolfgang, Ernst (1931), Die Kursbildung am Rentenmarkt, Stuttgart.